EBV-infection in cardiac and non-cardiac gastric adenocarcinomas is associated with promoter methylation of p16, p14 and APC, but not hMLH1

Epstein-Barr virus (EBV)-associated gastric carcinomas (GC) constitute a distinct clinicopathological entity of gastric cancer. In order to determine underlying distinct aberrant promoter methylation we tested cardiac and non-cardiac GC with regard to the presence of EBV. One hundred GC were tested by RNA- in situ hybridization for the presence of EBV by EBV-encoded small RNA (EBER). Aberrant promoter methylation was investigated by methylation-specific real-time PCR for p16, p14, APC and hMLH1. P16 protein expression was assessed by immunohistochemistry. In our selected study cohort, EBER-transcripts were detected in 19.6% (18/92) of GC. EBV-positive GC revealed significantly more often gene hypermethylation of p16, p14 and APC (p < 0.0001, p < 0.0001, and p = 0.02, respectively) than EBV-negative GC. The majority of GC with p16 hypermethylation showed a p16 protein loss (22/28). In contrast, no correlation between the presence of EBV and hMLH1 hypermethylation was found (p = 0.7). EBV-positive GC showed a trend towards non-cardiac location (p = 0.06) and lower stages (I/II) according to the WHO (p = 0.05). Hypermethylation of tumor suppressor genes is significantly more frequent in EBV-associated GC compared to EBV-negative GC. Our data add new insights to the role of EBV in gastric carcinogenesis and underline that EBV associated GC comprise a distinct molecular-pathologic as well as a distinct clinicopathological entity of GC.

underlying molecular mechanisms of EBV contributing to gastric carcinogenesis are more and more understood. Compared to EBV-negative GC, EBV-associated GC (EBVaGC) are characterized by distinct alterations on the molecular, trancriptional and translational level [1,[3][4][5][6][7][8][9][10][11]. In addition, the expression of the BARF1 oncogene of EBV has been identified in EBV-associated GC and has been reported to impact e.g. apoptotic or oncogenic gene expression [12][13][14][15][16]. EBV-positive GC have been shown to be associated with hypermethylation of the tumor suppressor genes p16, p14 and APC in Asian patients [7,17,18]. However, data on hypermethylation of these genes in Caucasian patients are very sparse [19]. The tumor suppressor genes p16, p14 and APC have been reported to be hypermethylated in GC at varying frequencies (range 10%-84%) [3,4,6,17,[20][21][22]. The hypermethylation of the DNA mismatch repair gene hMLH1 often occurs in a setting of microsatellite instability and was found only in EBV-negative GC so far [7]. In the present study we related the methylational status of p16, p14, APC and hMLH1 to the presence of EBV in GC. Since cardiac and non-cardiac GC (located in body or antrum) are characterized by distinct molecular alterations [23], 50 carcinomas from each tumor localization were analyzed. Further, EBV status was correlated to the available clinicopathological features, [24][25][26][27] in order to determine if aberrant promoter hypermethylation can be added to the list of distinct molecular changes of EBV-associated GC possibly contributing to EBVaGC as distinct clinicopathological entity.

Case selection and tissue microarrays
From a large ongoing study [23,28], we selected 50 patients with a carcinoma of the gastric cardia and 50 carcinomas, located in the distal stomach, namely in the body and antrum. Cardiac GC were defined as type II and type III carcinomas from the esophago-gastric junction, according to the classification of Siewert and Stein [29]. Type I carcinomas (Barrett carcinomas) were excluded. All tumors were adenocarcinomas, partly with signet ring histology. Details on application of WHO-classification and Lauren's classification have been described previously [30].

EBER-RNA in situ hybridization
The EBV status of the GC had been tested previously by EBER-RNA in situ hybridization (RISH) on tissue microarrays (TMA) [28]. The TMAs contained three 0.6 mm thick punches per tumor (Beecher Instruments, Silver Spring, MD, USA) and only for signet ring cell carcinomas six instead of three punches were used. One complete and tumor bearing core per case was regarded as sufficient for the evaluation of the EBV-status according to previous results that EBVassociated GC do not reveal heterogeneity with respect to EBER-transcript expression [13]. One of 50 cardiac carcinomas and 7 of 50 gastric carcinomas were excluded from further analyses due to technical reasons, e.g. loss of tissue cores during RISH. Thus, in 49 cardiac and 43 distal gastric carcinomas a specific result could be achieved (Table 1).

Methylation specific PCR
Details of methylation specific real-time PCR (MSP) for p16 INK4A , p14 ARF , APC, hMLH1 and control gene MYOD1 have been published previously [3,31]. In brief, genomic DNA from formalin-fixed, paraffin-embedded carcinoma tissue and corresponding tumor-free gastric smooth muscle tissue which served as constitutive negative control, were  [34]. Primer sequences and MSP conditions are summarized in Table 2.

Immunohistochemistry
The p16 protein expression was investigated in all cases with p16 hypermethylation by immunohistochemistry as published before [33,39]. Briefly, slides from one representative tumor block per case were stained using a monoclonal p16-antibody (Clone 16P07, Neomarkers, Westinghouse, CA, USA) at a dilution of 1:50 according to a standard avidin-biotin-peroxidase protocol. In 4 of 32 cases non-specific positivity was observed and these cases therefore were excluded.

Statistics
Statistical analyses were performed by the program SAS, version 9.1 [35]. Using the Χ 2 -test p-values<0.05 were considered as statistically significant. For analysis of age, patients were dichotomized before in two groups, younger or older than 60 years, respectively. For analysis of tumor diameter, patients were dichotomized before in two groups, less or more than 50 mm, respectively.
Further, we proved 15 of 58 (25.9%) carcinomas of a lower stage (I and II) to be EBV positive, but only 3 of 34 carcinomas in stages III and IV (8.8%), although this correlation marginally failed to reach statistical significance (p=0.05).
All results are summarized in Table 3.

Discussion
There are several indications, that aberrant methylation might be a central molecular mechanism of EBV infection in gastric carcinoma cells. When viral DNA enters a cell, it is often deactivated by methylation. This leads to repression of viral protein expression and escape of immune surveillance of cells with oncogenic potential. But the EBV induced methylation also affects the adjacent host DNA [21] and can even occur genome-wide [17]. This may result in inactivated tumor suppressor genes. Gene silencing by aberrant methylation (hypermethylation) of the tumor suppressor genes p16, p14, and APC and of hMLH1 are prevalent findings in GC [23,33,36]. In the present study, we proved that methylation of these tumor suppressor genes is an important mechanism in EBVaGC.
The actually epigentic silencing is demonstrated by p16 protein loss in most methylated cases.
These results have been confirmed by other authors, so far mostly in Asian patients [7,17,18]. Similar to our results, Chong and co-workers (2003) proved 88.2% of methylation in p14 and p16 each, which was significantly associated with EBV positivity [17]. These findings of high-density methylation strengthen the hypothesis that EBV infection causes a de novo and maintenance of methylation [22]. Similar to our results, Kang and co-workers found multiple genes, e.g. APC and p16 being hypermethylated in a frequency >90% in EBVaGC, defining a so-called methylator phenotype [7]. EBV-positive GC show a concordant methylation of multiple genes very much like to microsatellite instable (MSI) GC, although in our and other studies EBVaGC usually show no MSI and vice versa [7]. In our study EBV infection was not associated with hMLH1 hypermethylation, which is a marker for MSI. These results, which are corroborated by other investigators, indicate that EBV induces a malignant pathway, independent of MSI [21].
The different molecular pathways of EBV-positive GC are reflected by different genetic aberration. EBVassociated GC are characterized by distinct chromosomal aberrations and in addition distinct patterns of allelic losses, e.g. at the TP53 locus including p53 immunoreactivity, are associated with the EBV-status [5,10]. Furthermore, underlying the distinct molecular pathology, EBVassociated GC reveal a unique type of EBV-latency, which is mainly characterized by the expression of the viral oncogene BARF1 and the absence of LMP1 [13,37,38].
These molecular differences are also reflected by the finding that EBV-associated GC comprise a distinct clinicopathological entity: EBV-associated GC are linked to male predominance, lower age, intestinal histological type, proximal location, and a favorable prognosis [13,24,27].
The purpose of our study was not to investigate the clinico-pathological parameters associated with the prevalence of EBV, but the impact of differences in promoter methylations with regard to EBV.
However, the identification of the tight association to aberrant methylation of p16, p14 and APC in EBVaGC in this study, which was not observed to that extent in EBVnegative GC, strongly underline together with the missing association to hMLH1 hypermethylation that EBVaGC comprise a distinct clinicopathological entity.